rule-based detection of inconsistency in uml modelswl/papers/2002/uml02wl.slides.pdf · rule-based...
TRANSCRIPT
Rule-based Detection of Inconsistency in UML Models
WenQian LiuUniversity of Toronto
October 1, 2002
http://www.cs.utoronto.ca/~wl/papers/index.html
10/02/2002 WenQian Liu, University of Toronto 2
Outline
l The Problem l inconsistency management in design
models
l The Solution l knowledge base solution: production
systems
l Discussion and Summary
10/02/2002 WenQian Liu, University of Toronto 3
The Problem
lMaintaining consistency in large evolving UML design models is hard.
l Change introduces inconsistency, i.e.l conflicting description of the systeml violation of predefined constraints
10/02/2002 WenQian Liu, University of Toronto 4
Motivation
lManual inconsistency management is tedious and error prone.
lMany existing solutions use preventive and batch approaches.
lWe need computer assistance that can l react to change,l work incrementally, andl provide non-interruptive but timely
feedback.
10/02/2002 WenQian Liu, University of Toronto 5
Example 1 Redundancy
Sequence diagram 1: Request a new meetingSequence diagram 1: Request a new meeting Sequence diagram 2: Request a specific meetingSequence diagram 2: Request a specific meeting
10/02/2002 WenQian Liu, University of Toronto 6
Example 2
Law of Demeter
class Borrower { ···Book[] p = getLibrarian().findBookByCallNumber().listCopies();
···}
Conformance to Common Design Practices
10/02/2002 WenQian Liu, University of Toronto 7
Classes of Design Inconsistency
Design Consistency
Redundancy Conformance Change
Design Related
Data Related
Constraints
Standards
Edit blocks
10/02/2002 WenQian Liu, University of Toronto 8
Outline
l The Probleml The Solution l knowledge base solution: production
systems
l Discussion and Summary
10/02/2002 WenQian Liu, University of Toronto 9
Production Systems (1)
l Knowledge base (contains facts)l Rule base (contains rules)l Production system constantly
applies rules to factsl Rules modify facts
10/02/2002 WenQian Liu, University of Toronto 10
Production Systems (2)l Facts
l definition of a fact(type attribute:a1 … attribute:an)
l examples(transaction type:”debit” amount:50 accountid:00641)(account id:00641 balance:395)
l Rulesl definition
IF condition THEN actionl examples
IF (transaction type:”debit” amount:x accountid:a)(account id:a balance:y∧{≥x})
THEN REMOVE 1MODIFY 2 (balance [y-x])
l Basic operation:1. recognize2. resolve conflict3. act
10/02/2002 WenQian Liu, University of Toronto 11
Production Systems (3)
l reactive, incremental, extensiblel suitable for open-ended tasks l good for justification and explanation
10/02/2002 WenQian Liu, University of Toronto 12
Inconsistency Identification and Resolution
Four types of production rules 1. Inconsistency
l identify violations of consistency properties
2. Resolutionl resolve inconsistency upon receiving user’s choice
3. Cleanupl remove invalid inconsistency messages
4. Dynamic controll modify rule behaviors on the fly
10/02/2002 WenQian Liu, University of Toronto 13
Inconsistency Rule – Example 2
IF IF (sequenceMessage id:m1 from:T1 to:T2 return:c pid:p)(sequenceObject name:c type:T3∧{≠T2} pid:p)(sequenceMessage id:m2 from:T1 to:T3 pid:p)
THEN THEN ADD (inconsistency id:[newId()]∧s ruleid:”sc-1”location:((sequenceMessage m1)
(sequenceMessage m2)(sequenceObject c))
msg:”Violation of the Law of Demeter.”)ADD (userchoice id:[newId()] pid:s action:remove
targetID:m1 targetType:sequenceMessage)ADD (userchoice id:[newId()] pid:s action:remove
targetID:m2 targetType:sequenceMessage)
10/02/2002 WenQian Liu, University of Toronto 14
Resolution Rule – Example 2
IFIF (inconsistency id:s)(userchoice pid:s action:remove targetID:m
targetType:sequenceMessage)(userinput pid:s action:remove targetID:m)(sequenceMessage id:m)
THENTHEN REMOVE 1REMOVE 2REMOVE 3REMOVE 4
10/02/2002 WenQian Liu, University of Toronto 15
RIDE System Boundary
Synchronizer
Rule Editor Rule Engine Abstraction Layer
User
User
Jess RuleEngine
UML ModelEditor Legend
Component
System
Dependency
Implementation
10/02/2002 WenQian Liu, University of Toronto 16
Outline
l The Probleml The Solution l Discussion and Summary
10/02/2002 WenQian Liu, University of Toronto 17
Discussion
l worst-case complexity O(W2C-1)where W = size of the WM, C = number of patterns in a rule
l declarative (e.g. xlinkit, design guidance)l procedural (e.g. Argo/UML)l classes of consistency
10/02/2002 WenQian Liu, University of Toronto 18
Summary
l Contributionsl inconsistency classification
l redundancy, conformance to constraints and standards, changel rule-based detection and resolution approach
l incremental inconsistency detection and monitoringl single or multi-step resolution
l integration with UML editors
l Future Workl classification scheme and standard solutionsl verification of semantics over extensions to UMLl analysis of inconsistency patterns over history of edits
http://www.cs.utoronto.ca/~wl/papers/index.html
Questions?